--- title: "Advertising in ChatGPT for agencies: a channel that clears policy before it publishes" description: "A client ad account by key, a policy check before publishing, campaigns published paused, geo required, and CRM conversions in the shared summary." canonical: https://adsbeast.pro/platforms/en/chatgpt-ads language: en section: platforms product_route: https://adsbeast.pro/chatgpt-ads translations: - ru: https://adsbeast.pro/platforms/ru/chatgpt-ads - uk: https://adsbeast.pro/platforms/uk/chatgpt-ads - sk: https://adsbeast.pro/platforms/sk/chatgpt-ads - es: https://adsbeast.pro/platforms/es/chatgpt-ads - he: https://adsbeast.pro/platforms/he/chatgpt-ads - ar: https://adsbeast.pro/platforms/ar/chatgpt-ads --- # Advertising in ChatGPT for agencies: a channel that clears policy before it publishes ADS Beast runs advertising inside ChatGPT conversations as an ordinary media channel: the client ad account connects by key, every campaign is read against platform policy before it is published, it is published paused, and it is switched on only once the ads have been approved. Spend and results sit in the same summary as every other platform. ## What gets connected - **Connection**: A key issued inside the client ad account itself; every client has an account of their own - **What is read**: Campaigns, ad groups, ads, review status, spend and results in the shared summary - **What is applied**: Publishing paused; activation only after the ads are approved and a person confirms - **Targeting**: Geo is required, and conversation topics sit at ad group level in place of keywords - **Conversions**: Leads and deals from the CRM go to the platform Conversions API - **Who sees it**: People on the free plans of the platform; paid plans do not see advertising Creating the ad account, passing advertiser verification and attaching a card are done by the account owner on the platform side. The system takes over once the key exists. ## What the system does in the Advertising in ChatGPT channel ### Policy is read before publishing, not after a block The platform reviews more than the ad text: it looks at the advertiser, the domain, the landing page and where the visitor goes next. A neutral banner over a prohibited product does not pass and can cost you the account. So the client category and the prohibited topics are worked out before the campaign ever leaves for the platform. - A policy verdict with its reason: clear, needs a manual review, prohibited - A check that geo is set: without it the campaign would target every region - A warning about one-word conversation topics - that is keyword thinking, not intent ### The campaign is built from a description of the business An operator describes the client business in plain words and the system proposes a structure: ad groups, conversation topics, ad text and landing pages. Every part can be edited before publishing. The finished draft goes through the same policy check. - A campaign draft with ad groups, topics and ads - A setup note for each ad group - Edits to the draft before it is published ### Published paused, switched on in order The campaign, its ad groups and its ads are created paused, and the system then polls review status. Switching on is blocked while no ad is approved, and it runs bottom up: ads, ad groups, campaign. Every publishing step is idempotent, so a repeat run does not create a second object. - A paused campaign on the platform with a review mark on every ad - Activation only after a person confirms - A publishing log with the object identifiers on the platform ### CRM conversions and figures in the shared summary Leads and deals from the CRM go to the platform Conversions API every 30 minutes, in the same run as Meta and Google. Spend and results live in the shared tables, so analytics, alerts and automation rules see the channel without any extra work. - A spend line for the channel beside the client's other platforms - CRM conversions in the Conversions API - Alerts and automation rules shared with every other channel ## From the account key to the first impression The order is strict and it does not get shortened: check, publish paused, review, switch on. That way no money is spent before the links, the geo and the events have been checked. 1. **Connect the client account** - The account owner issues a key inside the account and hands it over. The system shows which capabilities the platform has already enabled for that account. 2. **Build the campaign and check it** - A draft from the business description, geo set, conversation topics, and a policy verdict. 3. **Publish it paused** - The campaign, ad groups and ads are created paused, and the system polls the review. 4. **Switch on after approval** - Once there are approved ads, a person confirms the activation. After that: figures and conversions in the shared summary. ## Figures from the code - 30 min - between CRM conversion uploads - 8 - channels in one summary, this one included - 15 min - between metric refreshes on the other platforms ## What the system will not do in the Advertising in ChatGPT channel - It will not create ad accounts, pass advertiser verification or attach cards - the account owner does that on the platform side. - It will not run every client through one ad account: that puts the whole portfolio behind a single block. - It will not publish campaigns already active, and it will not switch them on without approved ads. - It will not let a prohibited category through under neutral wording. ## Questions and answers ### How is an ad account key different from an ordinary platform key? They are different products. The advertising key is issued inside the ad account and serves that one account. An ordinary key for the models does not work for advertising. ### Who will see the ads in ChatGPT conversations? People on the free plans of the platform. Paid plans do not see advertising, so part of the paying business audience sits outside the inventory. That is worth allowing for in the plan. ### Why is geo required? Under the platform contract, a campaign with no geo can target every available region. The system blocks publishing without geo so the budget does not scatter across the world. ### Which categories will not pass? Adult content and dating, gambling, alcohol and tobacco, drugs, cryptocurrency, credit repair, politics, and consumer legal services. Finance and healthcare go through a manual review with a licence confirmed. ### Can the channel be run from the assistant? Yes: the assistant shows the state of the account, checks the draft against policy and publishes the campaign paused. It cannot switch a campaign on - a person does that. ## Connect a client account and read your first campaign against policy before publishing The first 14 days are free on any plan and you do not need a card to start. The channel gets the same analytics, alerts and approvals as every other platform. - Start free: https://adsbeast.pro/signup?lang=en - Pricing: https://adsbeast.pro/pricing?lang=en ## See also - [Rule checks: whether an ad will pass review, known before you send it](https://adsbeast.pro/features/en/compliance): Ads and landing pages read against platform rules and the limits of the client vertical before anything is sent, plus a post-mortem on rejections. - [Approvals: automation proposes, a person decides, the system applies](https://adsbeast.pro/features/en/approvals): Automation proposes changes to budgets, bids and statuses, a person confirms them, and the system applies them and reads the result back from the account. - [Visibility in assistant answers: are you named, and does the traffic arrive](https://adsbeast.pro/features/en/ai-visibility): Do assistants mention you in their answers: checked every 7 days, which crawlers visit the site, and how many people arrive from answers. - [Lead quality: who fills in the forms and how real they are](https://adsbeast.pro/features/en/leads): Leads from every platform and from the site in one feed with a bot score: flagged from 40 points, blocked from 70. Industry profiles, delivery checks.